DistatisFast {phylter}

Fast implementation the multivariate analysis method Distatis

Description

New implementation of the DISTATIS method for K matrices of dimension IxI. This version of Distatis is faster than the original one because only the minimum required number of eigenvalues and eigenvectors is calculated. The difference in speed is particularly visible when the number of matrices is large.

Usage

DistatisFast(matrices, factorskept = "auto", parallel = TRUE)

Arguments

matrices

A list of K distance matrices, all of the same dimension (IxI).

factorskept

Number of factors to keep for the computation of the factor. When "auto" (the default), a brokenstick model is used for the choice of the number of components to keep.

parallel

Should the matrix products computations be parallelized? Default to TRUE. scores of the observations.

Value

Returns a list containing:

• F: projected coordinates of species in the compromise

• PartialF: list of projected coordinates of species for each gene.

• alpha: array of length K of the weight associated to each matrix.

• lambda: array of length K of the normalization factors used for each matrix. lambda=1 always.

• RVmat: a KxK matrix with RV correlation coefficient computed between all pairs of matrices.

• compromise: an IxI matrix representing the best compromise between all matrices. This matrix is the weighted average of all K matrices, using 'alpha' as a weighting array.

• quality: the quality of the compromise. This value is between 0 and 1

• matrices.dblcent: matrices after double centering describes how much of the variance of the K matrices is captured by the compromise.

References

Abdi, H., Valentin, D., O'Toole, A.J., & Edelman, B. (2005). DISTATIS: The analysis of multiple distance matrices. Proceedings of the IEEE Computer Society: International Conference on Computer Vision and Pattern Recognition_. (San Diego, CA, USA). pp. 42-47.

Examples

# Get a list of matrices 
# from the carnivora dataset
data(carnivora) 
matrices <- phylter(carnivora, InitialOnly = TRUE, parallel = FALSE)$matrices

# Perform a Distatis analysis on these matrices: 
distatis <- DistatisFast(matrices, parallel = FALSE)

#distatis is a list with multiple elements: 
distatis$alpha #weigh of each matrix (how much it correlates with others)
distatis$RVmat #RV matrix: correlation of each matrix with each other
distatis$compromise # distance matrix with "average" pairwise distance between species in matrices
# etc.